A study by the Ellis Foundation in Alicante concludes that these ‘beautification’ algorithms whiten skin | The image returned by the screen can cause mental disorders or dysmorphia
The filters of Instagram, Tik Tok, and even those that we use in a video conference, whiten the user’s skin. Even the ones that are already white. For the “beauty canon” of these Artificial Intelligencesthe “beautified faces” are the white faces.
The filters they play against the diversity. They homogenize the faces in the same Caucasian canon. The algorithm equates being white with being beautiful. It lengthens the eyelashes, makes the eyes larger and elongated, the lips fuller, removes impurities from the skin, makes the nose small and the cheekbones angular. He does the same thing to everyone, no difference. No diversity.
These are some of the conclusions of the latest study of the Ellis Foundation of Alicante, a group of specialists focused on ethical research of Artificial Intelligence, focused on the relationship between people and intelligent systems. For the research, use has been made of databases with tens of thousands of faces to see the effects that the algorithms of these filters have on people.
Nuria Oliver is the Scientific Director of the Ellis Foundation and one of the promoters of this study. The center is currently dedicated to studying artificial intelligence applied to people so that have a positive social impact. In it, there are three lines of research, the first explores AIs that create models of human behavior, such as a city or a country: “the one we use to predict the progress of the pandemic is one of them,” explains Oliver.
The second line of research is the AI ”with which we interact” such as a chat or the applications that we use daily. And the last line and within which this study is developed is the “AI we trust”, which explores the ethical challenges posed by Artificial Intelligences.
live inside the filter
The conclusions of the study are worrisome since “There’s a demographic that won’t post any selfies or photos without a beautifying filter on,” Oliver explains. This, the homogenization of faces and a beauty canon unattainable for 99% of the population is the breeding ground for some mental disorders.
“On a psychological level this generates many problems for people. The first is the body dysmorphia, but also generates anxiety, eating disorders (ACT)and we even know of cases of young people who have undergone surgery to have a face similar to the Instagram filter,” Oliver denounces.
The expert qualifies that the beautification filters they are not bad per se and that in fact his foundation is always focused on seeking the positive social impact that Artificial Intelligence can have for people. May AI serve us, and not the other way around.
The first question asked in the study was the following: “The filters of instagram play against diversity?” When they applied them to the database of faces they realized that yes. “They all create the same similarities to beautify that face. Socially we should be concerned about this, because it’s killing diversity,” Oliver explains.
Another question asked is if the AI can identify us after applying the filter, that is, if our face changes so much as to continue being recognizable. The answer was also yes, we are still identifiable with the filter.
Racist beauty canon
Another question that was raised in the study was: “What canon of beauty does it codify?” First, a white racial bias. Oliver claims that, for these algorithms, ‘beautified faces’ are whiter. The algorithm is able to detect the person’s race; if she is Asian, black, Latino, white… But “it beautifies and whitens all of them, even white skins it makes them even whiter”, says Oliver.
In summary “The algorithm equates being white with being beautiful.. And this has tremendous historical connotations to do with colonialism,” he explains.
But the filter doesn’t do just that. It also lengthens the eyelashes, makes the eyes larger and elongated, the lips fuller, does not create impurities in the skin, makes the nose smaller and better cheekbones.. This, for Oliver, is especially interesting because “most filters are created by users, and everyone is going towards the same canon of beauty,” she says.
Taking all this into account, the psychological consequences are very difficult to avoid. “It is very difficult for a person to be abstracted if the image they project on the networks is very different from the one returned by the mirror. Even if you know that this is not real, it is difficult to separate yourself. Even if you know that it is not like that, you compare yourself. You also know that everyone uses filters, but your image causes you unhappinessOliver remarks.
Then there is a very relevant topic. The reality is that this content, photos full of beautification filters, is what generates more likes on Instagram, and therefore the one that best positions the algorithm because he sees that he is catching more attention. “Ultimately, people look for external validation, and if something isn’t embellished you almost certainly know it’s not going to succeed, so it’s ‘not valid,'” Oliver remarks.